target/riscv: define register printers

[en-sc]

This commit implements facilities to decode register fields for each register described in debug_defines.h.

Namely:

• define <REG_NAME>_FORMAT_STR
• macro <REG_NAME>_FORMAT_ARGS(value)

The use case is demonstarted by changing debug message in execute_abstract_command()

Change-Id: I63733d8d36385d89ca15de1a43139134bc488c4f

[en-sc]

This commit fails checkpatch, since the macro argument is reused.

CHECK:MACRO_ARG_REUSE: Macro argument reuse 'value' - possible side-effects?

IMHO, this is an insignificant drawback, since there are other macro which also reuse their's arguments (e.g. get_field)

If you have any suggestions on how to fix or disable (locally) this warning, I would be glad to hear them.

[en-sc]

The changes to debug_defines.h are generated by applying the patch riscv/riscv-debug-spec#858 on top of previously used version of riscv-debug-spec (d749752).

[en-sc]

debug_register_printers.c implements decoder for register values based on the new interface of debug_defines.h.

riscv_debug_reg_to_s() can be used to decode register value. If the
pointer to buffer is NULL it does not print anything, just returns the
length of the string.

The format is:
<register_value> { <field_name>=<field_value_name or field_value>, ..., }

e.g:

0x400382 { version=2, confstrptrvalid=invalid, hasresethaltreq=0, authbusy=ready, authenticated=true, anyhalted=1, allhalted=1, anyrunning=0, allrunning=0, anyunavail=0, allunavail=0, anynonexistent=0, allnonexistent=0, anyresumeack=0, allresumeack=0, anyhavereset=0, allhavereset=0, impebreak=1, stickyunavail=current, ndmresetpending=false, }

0x321009 { regno=0x1009, write=arg0, transfer=enabled, postexec=disabled, aarpostincrement=disabled, aarsize=64bit, cmdtype=0, }

[en-sc]

The only warning from checkpatch left is:

WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1
#27: FILE: src/target/riscv/debug_defines.c:1:

I'm not shure what is the best approach.
Should I leave it as-is (there is no license tag in debug_defines.h also) or should I add it?

[timsifive]

WARNING:SPDX_LICENSE_TAG: Missing or malformed SPDX-License-Identifier tag in line 1

We can ignore that warning and merge it anyway when the time comes.

[timsifive]

This code looks good. I'll head over to riscv/riscv-debug-spec#858 now.

[en-sc]

@timsifive, @JanMatCodasip, can this be merged (since riscv/riscv-debug-spec#858 is already merged), or should I make some adjustments?

[JanMatCodasip]

Despite having somewhat different opinion on what the API of debug_defines.h could be, I like this change overall and it looks to me like a nice step forward in unification of the log prints. Thanks!

I believe there are few minor items to address prior to the merge.

[JanMatCodasip]

Please, can you re-generate these files from a non-dirty revision of the debug spec repo.

[en-sc]

This is intentional. I would like to avoid updating the version of the spec used to generate this file, since it would require a lot of changes in code (some constants were renamed and so on). The version is marked as dirty to indicate that the generators were cherry-picked on top of d749752. If you insist, I will update debug_defines.h to current version, though I'm not sure if it is the best time to do so.

[JanMatCodasip]

I would like to avoid updating the version of the spec used to generate this file, since it would require a lot of changes in code (some constants were renamed and so on).

Understood. In that case, it is my recommendation to leave this merge request as is, but please make a proper upgrade of the debug_defines (from a clean commit) in a subsequent merge request.

The version is marked as dirty to indicate that the generators were cherry-picked on top of d749752

This is exactly the reason why I suggest to use a clean commit -- we can't expect other developers to figure out what needs to be cherry-picked where (what mixture of commits to make) in order to replicate your work, e.g. during troubleshooting.

Thanks.

[en-sc]

I have re-checked and the changes to the code are limited (only one function), so I updated the version of debug_defines.h in this MR.

[JanMatCodasip]

This looks fine to me.

I have only one remark #892 (comment) (which can be addressed as a separate merge request to keep the changes easier to review).

Thank you.

[en-sc]

@timsifive, @JanMatCodasip can this be merged or is there some action required to be done?